CI with `num_hidden_layers=2` 🚀🚀🚀 (#25266)

* CI with layers=2

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

tests/models/chinese_clip/test_modeling_chinese_clip.py

@@ -69,7 +69,7 @@ class ChineseCLIPTextModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",
@@ -246,7 +246,7 @@ class ChineseCLIPVisionModelTester:
         is_training=True,
         hidden_size=32,
         projection_dim=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,

tests/models/clap/test_modeling_clap.py

@@ -287,7 +287,7 @@ class ClapTextModelTester:
         vocab_size=99,
         hidden_size=32,
         projection_dim=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,

tests/models/clip/test_modeling_clip.py

@@ -86,7 +86,7 @@ class CLIPVisionModelTester:
         is_training=True,
         hidden_size=32,
         projection_dim=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,
@@ -261,7 +261,7 @@ class CLIPTextModelTester:
         vocab_size=99,
         hidden_size=32,
         projection_dim=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,

tests/models/clip/test_modeling_flax_clip.py

@@ -35,7 +35,7 @@ class FlaxCLIPVisionModelTester:
         num_channels=3,
         is_training=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,
@@ -252,7 +252,7 @@ class FlaxCLIPTextModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,

tests/models/clipseg/test_modeling_clipseg.py

@@ -78,7 +78,7 @@ class CLIPSegVisionModelTester:
         num_channels=3,
         is_training=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,
@@ -228,7 +228,7 @@ class CLIPSegTextModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,
@@ -346,7 +346,15 @@ class CLIPSegTextModelTest(ModelTesterMixin, unittest.TestCase):
 
 
 class CLIPSegModelTester:
-    def __init__(self, parent, text_kwargs=None, vision_kwargs=None, is_training=True):
+    def __init__(
+        self,
+        parent,
+        text_kwargs=None,
+        vision_kwargs=None,
+        is_training=True,
+        # This should respect the `num_hidden_layers` in `CLIPSegVisionModelTester`
+        extract_layers=(1,),
+    ):
         if text_kwargs is None:
             text_kwargs = {}
         if vision_kwargs is None:
@@ -356,6 +364,7 @@ class CLIPSegModelTester:
         self.text_model_tester = CLIPSegTextModelTester(parent, **text_kwargs)
         self.vision_model_tester = CLIPSegVisionModelTester(parent, **vision_kwargs)
         self.is_training = is_training
+        self.extract_layers = extract_layers
 
     def prepare_config_and_inputs(self):
         text_config, input_ids, attention_mask = self.text_model_tester.prepare_config_and_inputs()
@@ -371,7 +380,7 @@ class CLIPSegModelTester:
             self.vision_model_tester.get_config(),
             projection_dim=64,
             reduce_dim=32,
-            extract_layers=[1, 2, 3],
+            extract_layers=self.extract_layers,
         )
 
     def create_and_check_model(self, config, input_ids, attention_mask, pixel_values):

tests/models/codegen/test_modeling_codegen.py

@@ -47,7 +47,7 @@ class CodeGenModelTester:
         vocab_size=256,
         hidden_size=32,
         rotary_dim=4,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/convbert/test_modeling_convbert.py

@@ -53,7 +53,7 @@ class ConvBertModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/cpmant/test_modeling_cpmant.py

@@ -49,7 +49,7 @@ class CpmAntModelTester:
         use_mc_token_ids=False,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=3,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         num_buckets=32,

tests/models/ctrl/test_modeling_ctrl.py

@@ -49,7 +49,7 @@ class CTRLModelTester:
         use_mc_token_ids=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/data2vec/test_modeling_data2vec_audio.py

@@ -59,7 +59,7 @@ class Data2VecAudioModelTester:
         conv_bias=False,
         num_conv_pos_embeddings=16,
         num_conv_pos_embedding_groups=2,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         hidden_dropout_prob=0.1,
         intermediate_size=20,

tests/models/data2vec/test_modeling_data2vec_text.py

@@ -57,7 +57,7 @@ class Data2VecTextModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/data2vec/test_modeling_data2vec_vision.py

@@ -59,7 +59,7 @@ class Data2VecVisionModelTester:
         is_training=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/deberta/test_modeling_deberta.py

@@ -47,7 +47,7 @@ class DebertaModelTester(object):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/deberta_v2/test_modeling_deberta_v2.py

@@ -48,7 +48,7 @@ class DebertaV2ModelTester(object):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/deit/test_modeling_deit.py

@@ -69,7 +69,7 @@ class DeiTModelTester:
         is_training=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/dinov2/test_modeling_dinov2.py

@@ -57,7 +57,7 @@ class Dinov2ModelTester:
         is_training=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/distilbert/test_modeling_distilbert.py

@@ -50,7 +50,7 @@ class DistilBertModelTester(object):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/distilbert/test_modeling_flax_distilbert.py

@@ -47,7 +47,7 @@ class FlaxDistilBertModelTester(unittest.TestCase):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/dpr/test_modeling_dpr.py

@@ -48,7 +48,7 @@ class DPRModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/dpt/test_modeling_dpt.py

@@ -53,7 +53,7 @@ class DPTModelTester:
         is_training=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         backbone_out_indices=[0, 1, 2, 3],
         num_attention_heads=4,
         intermediate_size=37,
@@ -62,7 +62,7 @@ class DPTModelTester:
         attention_probs_dropout_prob=0.1,
         initializer_range=0.02,
         num_labels=3,
-        neck_hidden_sizes=[16, 16, 32, 32],
+        neck_hidden_sizes=[16, 32],
         is_hybrid=False,
         scope=None,
     ):